1 Jan 2006

Sun, 1 Jan 2006 00:43:26 GMT

I really wanted to reply to the 50 BEST Robots thread, but I dont see a way to do that quite yet, so I'll post my response here.

Let me first state the my views expressed here are of my own only, but… Dont even get me started on DARPAs Grand Challenge this past year. While I respect those who participated, Im just so disappointed with how 'dumb down' the game challenge had to become to get people to actually finish the race. Take the spread two years ago. Let's call them 1 st Gen Teams. The best team (CMUs team) made it like 7 miles and their Hummer ultimately caught fire after it high centered. Everyone else embarrassed the robotics world as they, collectively, barley made it off the starting line. Where was this super high tech robotics community that is so syndicated? Fast forward to last years teams, appropriately we'll call them 2 nd Gen Teams. Five teams finished the 130 some odd mile trek and four finished under the 10 hour time limit. What???Huh? What massive developments in technology allowed these miraculous changes for people to finish? One might argue that the added year of testing is enough to justify the better finishes. I'm not here to downplay the value of testing. Trust me, I understand testing and the benefits therein, but let's be realistic. The total miles completed by all the 1 st Gen teams, and let's be liberal, totaled prolly less than 10 miles. Total miles completed by 2nd Gen teams, and let's be conservative, was probably around 1000 miles. One year's worth of testing accounted for 100x in the amount of miles collectively completed by the teams? In the spirit of the season, 'Bah-Humbug.' I don't think so. Enter media pressure to the show. The media had hyped up the DARPA Grand Challenge so much that after an abysmal showing of robotic contestants, the game was alternated to have more success stories. Let's look at a little known fact. The day of the race, teams were literally allowed to download a waypoint map (over 3000 points I might add) from the starting location to the final destination. So now the only thing 2 nd Gen teams had to do was automate a truck (pah-leeze) and close a position loop around a GPS signal and have it avoid some obstacles for 60 meters at a time. Effin lame if you ask me. And now this is worth recognition as one of the greatest robotic achievements of the 21 st century? Since when does DARPA sell out to get good press? Since when does DARPA reduce the scope of a challenge to make it easier for people to do? DARPA prides themselves on never taking on a task if there is a higher than 10% success rate. They only attempt the really, really hard problems? Give me an effin break. What surprises me is that everyone didn't finish. That just reinforces how freakin lame most of the robotic community actually is when it comes to mobile navigation. If Im not mistaken, 1st Gen Teams weren't given much data and only told to head 'That-away.' And, as empirical evidence showed, teams could hardly automate a truck let alone head 'that-away'. What is more depressing is this is being hailed in the media as such a novel achievement. Has it always been this way? Have most of our amazing achievements only been a shred of what they were hyped to be? I mean, I think we landed on the moon. What I'm really tired of is reading technology magazines, and now having some insight into the projects they write about, knowing how 'off' they actually are. When are we gonna fess up and say we ain't really as advanced in robotics as we think we are and start owning up to it. These habitual oversights and denials will continue to push us further and further behind real roboticists, like say the Japanese.

I dont want to sound like I dont respect the work that went into finishing the DARPAs challenge, believe me I do. I just dont want to shy away from the really hard problems that we face today in robotics, like true autonomous mobile navigation. Ask me to get from Calitown, USA to Vegas and I can sure as hell bet you I wouldnt require a GPS point every 60 meters.

robots.net blog for burnt_circuits

1 Jan 2006